Dzulfidho Wijianto Putra, Ahmad Fahrudi Setiawan, Nurlaily Vendyansyah

Peramalan Curah Hujan Menggunakan Metode Holt-Winters Exponential Smoothing

Introduction

Peramalan curah hujan menggunakan metode holt-winters exponential smoothing. Ramalkan curah hujan bulanan dengan Holt-Winters Exponential Smoothing. Evaluasi akurasi (MAE, RMSE) untuk dukungan keputusan sektor pertanian dan manajemen air.

Abstract

Rainfall is a crucial climatological parameter for agriculture, tourism, and water resource management. Its seasonal and fluctuating nature requires accurate forecasting methods to capture historical patterns. This study forecasts monthly rainfall using data from Ngaglik, Temas, and Tinjumoyo stations between January 2021 and December 2024, totaling 48 observations. The Holt–Winters Exponential Smoothing Additive method was chosen due to stable annual seasonal patterns. Model accuracy was assessed with Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). Results show varying optimal parameters across stations. Ngaglik achieved the best performance with α = 0, β = 0, γ = 0.81, yielding MAE 64.39 mm and RMSE 90.84 mm. Temas recorded MAE 69.25 mm and RMSE 102.19 mm with γ = 0.78, while Tinjumoyo produced MAE 73.95 mm and RMSE 109.42 mm with γ = 0.56. This study highlights the effectiveness of Holt–Winters Additive forecasting and provides accuracy evaluations to support data-driven decisions in rainfall-dependent sectors.

Review

This study addresses a critically important topic: the accurate forecasting of rainfall, a climatological parameter vital for sectors such as agriculture, tourism, and water resource management. Recognizing the seasonal and fluctuating nature of rainfall, the authors employ the Holt–Winters Exponential Smoothing Additive method to predict monthly rainfall using a four-year dataset (January 2021 to December 2024) from three distinct stations: Ngaglik, Temas, and Tinjumoyo. The selection of the Additive method is explicitly justified by the assumption of stable annual seasonal patterns, which is a key prerequisite for its effective application. The clear objective and well-defined methodology lay a solid foundation for the research. The paper effectively demonstrates the application and utility of the Holt–Winters Additive model, evaluating its performance using established metrics: Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). A significant strength is the presentation of station-specific optimal parameters and corresponding error metrics, which highlights the localized nature of meteorological patterns and the tailored approach needed for accurate forecasting. Ngaglik station, for instance, achieved the best performance with an MAE of 64.39 mm and RMSE of 90.84 mm. The varying gamma (γ) values across stations (0.81, 0.78, 0.56) underscore the importance of parameter tuning for each location, ultimately providing actionable accuracy evaluations to support data-driven decisions in rainfall-dependent sectors. While the study successfully applies the Holt–Winters Additive method, its contribution could be further enhanced by incorporating a comparative analysis with alternative forecasting models, such as ARIMA, SARIMA, or even more complex machine learning approaches, to firmly establish its superior performance or delineate specific niches where it excels. Additionally, providing context for the reported MAE and RMSE values—perhaps by relating them to the average monthly rainfall in the respective regions or discussing the practical implications of these error margins for the targeted applications—would significantly improve the interpretability for practitioners. Finally, a brief discussion on the robustness of the model in the face of climate variability or extreme weather events, and a more detailed explanation of the parameter optimization process, would add further depth and transparency to this otherwise valuable research.

Full Text

Comments

You need to be logged in to post a comment.